36x24 digital!
2001-11-08 16:48:26+00 by
Dan Lyke
4 comments
The B&H digital catalog arrived, with lots of cool toys that I can't possibly justify. But, interestingly, the Mega-Vision S3 and the Fuji Luma are 36x24mm 3k by 2k CCD backs for medium format SLRs. This is the holy grail of 35mm photography, equivalent resolution to better films, with CCD sensitivity and grain, so although it costs $15 to $20k right now and is limited to the medium format, it's just a matter of waiting.
[Meta snicker: The categorizer thought California Culture was an appropriate tag for this entry. I think it keyed off "sensitivity"...]
[ related topics:
Photography
]
comments in ascending chronological order (reverse):
#Comment made: 2002-02-21 05:33:15+00 by:
TC
Ooops my bad! It keyed off "cool". I suppose cool is no longer excluesively Californian. Bad human! I'll go fix it.
#Comment made: 2002-02-21 05:33:15+00 by:
Dan Lyke
Speaking of which, I need to run that little script to update categories for older entries, then find a real solution to the issue...
#Comment made: 2002-02-21 05:33:16+00 by:
John Abbe
Automated categorizer? Someone say more?
#Comment made: 2002-02-21 05:33:16+00 by:
Dan Lyke
We have a database of topics and keywords. When a new entry gets made, the keywords are scanned and the entry is added to those topics, and then on the editing page we can uncheck those categories and add new ones (even create a new one on the fly).
Not all of us edit our categorizations, but having a list to remove from, rather than looking through a pulldown with all the topics means that the categorizer gets used.
Of course the keyword system gets confused, often with hilarious results.
To do:
- Newly added categories don't get run against previous entries,
so older entries don't even have wrong categories.
- I keep promising to do a "phrase which indicates this was a bad
choice" or "phrase which would have selected this" next to the
checkboxes and pulldowns so that we can build a negative
keyword/phrase system and build a larger positive one without
having to go to the keyword editing page.
- I've been playing with some simple template based English
parsing on a couple of obituaries and "on this day" web pages
for the past year, trying to derive easily expressable facts
("Megan Parlen" is an "Actress" tied to entity "Hang Time"
and was born on 1980-07-09, to take a random one). I haven't
yet written the learning bits of this, but obviously a little
fuzzy matching will give me certainties that occupations and
traits attached to these sentences are those, rather than
additional names. So eventually I'll have something learning
simple sentence structure, and if I find a sentence that
matches that structure I can use that information in the
categorizer.
The nice thing about the category experiment is that it's got a feedback loop, and miscategorizations actually make it more interesting.